Message-ID: <1830373312.9459322.1611241025034.JavaMail.zimbra@mnhn.fr>
Date: 2021-01-21T14:57:05Z
From: Marianne DEBUE
Subject: [R-meta] effect size calculation with a null standard deviation
Hi everyone,
I'm conducting a meta-analysis in ecology. I'm using Morris "dpcc1" formulas ( [ https://journals.sagepub.com/doi/10.1177/1094428106291059 | https://journals.sagepub.com/doi/10.1177/1094428106291059 ] ) to calculate the effect size and its variance.
The effect size calculation implies a difference between post- and pre-Mean which is then divided by the pre-Standard deviation.
I was wondering how to deal with studies which have pre-Standard deviation = 0 (leading to a division by 0) ?
How to deal with studies which have pre-Mean = post-Mean and a pre-Standard deviation = 0 (leading to 0 divided by 0) ? If pre-Mean = post-Mean , can we consider that the effect size is null, whatever the pre-Standard deviation ?
The variance calculation implies a division by (N - 3) (N : sample size).
How to deal with studies which have N = 3 ( leading to a division by 0) ?
Thank you for your help,
Marianne
[[alternative HTML version deleted]]